Stress and Interferon Signalling-Mediated Apoptosis Contributes to Pleiotropic Anticancer Responses Induced by Targeting NGLY1

www.nature.com/bjc

ARTICLE Cellular and Molecular Biology Stress and interferon signalling-mediated apoptosis contributes to pleiotropic anticancer responses induced by targeting NGLY1

Ashwini Zolekar1, Victor. J. T. Lin1, Nigam M. Mishra1, Yin Ying Ho2, Hamed S. Hayatshahi1, Abhishek Parab3, Rohit Sampat1, Xiaoyan Liao4,6, Peter Hoffmann2,5, Jin Liu1, Kyle A. Emmitte1 and Yu-Chieh Wang1

BACKGROUND: Although NGLY1 is known as a pivotal enzyme that catalyses the deglycosylation of denatured glycoproteins, information regarding the responses of human cancer and normal cells to NGLY1 suppression is limited. METHODS: We examined how NGLY1 expression affects viability, tumour growth, and responses to therapeutic agents in melanoma cells and an animal model. Molecular mechanisms contributing to NGLY1 suppression-induced anticancer responses were revealed by systems biology and chemical biology studies. Using computational and medicinal chemistry-assisted approaches, we established novel NGLY1-inhibitory small molecules. RESULTS: Compared with normal cells, NGLY1 was upregulated in melanoma cell lines and patient tumours. NGLY1 knockdown caused melanoma cell death and tumour growth retardation. Targeting NGLY1 induced pleiotropic responses, predominantly stress signalling-associated apoptosis and cytokine surges, which synergise with the anti-melanoma activity of chemotherapy and targeted therapy agents. Pharmacological and molecular biology tools that inactivate NGLY1 elicited highly similar responses in melanoma cells. Unlike normal cells, melanoma cells presented distinct responses and high vulnerability to NGLY1 suppression. CONCLUSION: Our work demonstrated the signiﬁcance of NGLY1 in melanoma cells, provided mechanistic insights into how NGLY1 inactivation leads to eradication of melanoma with limited impact on normal cells, and suggested that targeting NGLY1 represents a novel anti-melanoma strategy.

British Journal of Cancer (2018) 119:1538–1551; https://doi.org/10.1038/s41416-018-0265-9

BACKGROUND congenital deglycosylation disorder, were recently identified.2,5,8,9 As a pivotal glycosidase known for catalysing the removal of Many of these mutations cause premature termination of glycans from N-glycosylated asparagine residues, NGLY1 (a.k.a. translation, leading to complete loss of NGLY1 in the patients. PNGase) enables the deglycosylation of denatured glycoproteins Until this discovery, the disease implications of NGLY1 had not and allows proteasome-mediated protein degradation to effi- been definitive. ciently occur.1–5 A TGase-superfamily (PNGase-core) domain exists NGLY1 deficiency causes a broad spectrum of disease in NGLY1 proteins ranging from yeast to human,4 suggesting the phenotypes with incomplete penetrance in patients.2,5,8,9 Many evolutionarily conserved significance of NGLY1 enzymatic activity NGLY1 deficiency-associated phenotypes are closely related to in cells. It is known that loss of NGLY1 function in cells can cause developmental delay and congenital abnormalities, suggesting the accumulation of aberrant proteins in the cytosol and the the significant role and intricate regulation of this glycosidase in interruption of endoplasmic reticulum-associated protein degra- the normal development of human organs. Despite the recently dation (ERAD).1,4,5 Therefore, NGLY1 defects are likely to affect the gained knowledge about NGLY1 deficiency, there is limited quality control and homeostasis of many cellular proteins, information regarding the responses of human cancer cells and subsequently perturbing signalling pathways, cell physiology, terminally differentiated somatic cells to NGLY1 suppression. and organ development. The studies of an NGLY1 ortholog gene, NGLY1 is commonly expressed in many types of normal and PNGase-like (Pngl), in fruit fly and fungus also indicate that NGLY1 cancer cells (www.proteinatlas.org),10 suggesting that NGLY1 could be involved in the regulation of cell normality through an could be essential for a variety of human cells regardless of their enzymatic activity-independent mechanism.6,7 Interestingly, pathophysiological conditions. Notably, NGLY1 appears to be human NGLY1 gene mutations that result in NGLY1 deficiency, a highly expressed in certain human cancer cells (e.g., melanoma

1Department of Pharmaceutical Sciences, UNT System College of Pharmacy, University of North Texas Health Science Center, Fort Worth, TX, USA; 2Adelaide Proteomics Centre, The University of Adelaide, Adelaide, Australia; 3Department of Mathematics, Purdue University, West Lafayette, Indiana, USA; 4Department of Pathology, University of California, San Diego, San Diego, CA, USA and 5Future Industries Institute, University of South Australia, Adelaide, Australia Correspondence: Y-C. Wang ([email protected]) 6Present address: Department of Pathology and Laboratory Medicine, University of Rochester Medical Center, Rochester, NY, USA These authors contributed equally: Ashwini Zolekar, Victor J.T. Lin. Received: 26 April 2018 Revised: 11 August 2018 Accepted: 31 August 2018 Published online: 2 November 2018

© The Authors 2018 Published by Springer Nature on behalf of Cancer Research UK Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1539 and ovarian cancer), while low-to-undetectable in their normal Knockdown of NGLY1 and GADD153 counterpart tissues (e.g., skin and ovary) (www.proteinatlas.org).10 The knockdown of NGLY1 expression in melanoma cells was These observations raise an intriguing possibility that NGLY1 may achieved by the transduction of pZIP-TRE3GS lentiviral expression be crucial for cancer development and progression. Moreover, vectors that carry two independent shRNA sequences (Supple- cancer cells, at least in certain cancer types, may be particularly mentary Materials and Methods; TransOMIC Technologies, Hunts- vulnerable to loss of NGLY1 compared with normal cells. ville, AL). A pZIP-TRE3GS vector that carries a NT-shRNA sequence In this study, we systematically and mechanistically examined was used as the control. The expression of the shRNA sequences the significance of NGLY1 in melanoma cells and how it may be and an open reading frame of the ZsGreen reporter is driven exploited as a novel anticancer target. In addition, we developed by the TRE3GS doxycycline-inducible promoter. The transduced and tested novel covalent inhibitors that suppress NGLY1 activity cells were selected using puromycin for a prolonged period in human cells. Our results strongly support that (~4 weeks) to obtain the stable clones of cancer cells that carry NGLY1 suppression in melanoma cells elicits multifaceted inducible NT-shRNA, NGLY1-shRNA645 and NGLY1-shRNA647 cancer-elimination responses, and that targeting NGLY1 and sequences. protein deglycosylation may represent a novel anticancer strategy The knockdown of GADD153 expression in melanoma cells with the opportunity for a broad therapeutic window. was achieved by the transduction of pZIP-hEF1a-RFP lentiviral expression vectors that carry three independent shRNA sequences (Supplementary Materials and Methods; TransOMIC MATERIALS AND METHODS Technologies, Huntsville, AL). A pZIP-hEF1a-RFP lentiviral Cell Culture expression vector carries a NT-shRNA sequence was used as Human dermal fibroblasts were cultured in DMEM (Thermo Fisher the control. The expression of the shRNA sequences and an Scientific, Carlsbad, CA) containing 10% fetal bovine serum (FBS; open reading frame of the RFP reporter is driven by the human Thermo Fisher Scientific, Carlsbad, CA) at 37°C. HEMl and HEMd EF1α gene promoter. (ScienCell Research Laboratories, Carlsbad, CA) cells were cultured in melanocyte medium (MelM; ScienCell Research Laboratories, Overexpression of human NGLY1 Carlsbad, CA). Human melanoma cells were cultured using RPMI- A pLenti expression vector that carries a Myc-DDK-tagged-human 1640 medium (Thermo Fisher Scientific, Carlsbad, CA) or DMEM/ NGLY1 open reading frame driven by a CMV promoter (OriGene F12 medium (Thermo Fisher Scientific, Carlsbad, CA) containing Technologies, Rockville, MD) was transduced into cells for the

1234567890();,: 10% FBS. WA09 human embryonic stem cells (hESCs) were overexpression of NGLY1. A pLenti-C-Myc-DDK empty vector was obtained from the WiCell Stem Cell Bank (WiCell Research used as the transduction control. Institute, Madison, WI). HMi-50611 and NGLY1Pt1i-509 hiPSCs were established using CytoTune Sendai Reprogramming Kit Immunohistochemistry (IHC) and Fluorescence Staining (Thermo Fisher Scientific, Carlsbad, CA). We followed the The general procedure for antibody-mediated fluorescence previously described method12 for culturing undifferentiated staining was previously described12 and provided as part human pluripotent stem cells (hPSCs) in a feeder cell-free of Supplementary Materials and Methods. The detailed informa- condition, except the use of TeSR-E8 medium (Stemcell Technol- tion of primary antibodies was summarised in Supplementary ogies, Vancouver, Canada) and L7 hPSC passaging solution (Lonza, Table S2. Walkersville, MD) in this study. The detailed information of cells used in this study was summarised in Supplementary Table S1. Immunoblotting The experiments using hPSCs were performed in compliance with The general procedure for immunoblotting was described in a the guidelines and approval of the institutional biosafety previously published report,13 except that cell lysates were committee at UNTHSC. All cells were periodically tested using prepared using M-PER mammalian protein extraction reagent the MycoAlert mycoplasma detection kit (Lonza, Walkersville, MD) (Thermo Fisher Scientific, Carlsbad, CA) containing EDTA-free and free of mycoplasma. protease inhibitor and phosphatase inhibitor cocktails (Millipore Sigma, St. Louis, MO). The detailed information of primary Melanoma Patient Samples antibodies was summarised in Supplementary Table S2. HRP- The RNA samples of tumour tissues from randomly selected conjugated secondary antibodies were from Jackson ImmunoR- melanoma patients were obtained from OriGene Technologies esearch Laboratories (West Grove, PA). (Rockville, MD). The tissue arrays that contain 8 cases of human normal skin and 36 cases of melanoma tumours were acquired Flow Cytometry from BioChain Institute (Newark, CA). The procedures were provided as part of Supplementary Materials and Methods. CRISPR-Cas9-mediated Gene Editing For CRISPR-Cas9-mediated gene editing to knockout the expression Cell Viability Test of NGLY1 in hPSCs, we designed two NGLY1-targeting sgRNA The procedures were provided as part of Supplementary Materials sequences (sgRNA37:5’CATTCAACAGCTCCTCTGAC3’ and sgRNA39:5’- and Methods. GATCTGATGACTGCCCTTGA3’)usingtheCRISPRDesignTool(http:// crispr.mit.edu/). These two sgRNA sequences were independently Gene Expression Analysis by qRT-PCR and Microarrays cloned into a lentiCRISPRv2 plasmid (Addgene, Cambridge, MA) to The procedures for microarray analysis were provided as part generate two constructs of an one-vector system for sgRNA and of Supplementary Materials and Methods. The test of cellular Cas9 expression. WA09 hESCs transduced with the sgRNA and Cas9 pluripotency based on the transcriptomic features of cell samples expression constructs were selected using puromycin and sub- was performed using the PluriTest (http://pluritest.org/).14 Multiplex jected to a single-cell cloning process. Using a surveyor mutation qRT-PCR was performed using cDNA generated from the RNA detection kit (Integrated DNA Technologies, Coralville, IA) to samples and Taqman® assays for the NGLY1, FABP7, RSAD2, CCL5, examine indel mutations at the editing sites followed by western IFNB1 and ACTB (internal control) genes (assay ID# Hs01046153_m1, blotting to test NGLY1 expression, hESCs with NGLY1 gene Hs00361424_g1, Hs00369813_m1, Hs00982282_m1, Hs01077958_s1 mutations that lead to the ablation of NGLY1 expression were and Hs03023943_g1; Thermo Fisher Scientific, Carlsbad, CA), chosen and further expanded. according to the manufacturer’s instructions. Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1540

a 8 bc 2 Normal Melanoma 27 NGLY1 * 26 25

2 * HEMI HMi-506 UACC257 COLO829 MALME3M SK-MEL-2 SK-MEL-5 451Lu MEL1617 (KDa) 23 75 NGLY1 22 50 21 pMEK1/2 0 2 50 2–1 MEK1/2

Relative normalized expression Relative –2 2 50 pERK1/2 2–3

–4 50 2 Tumor Normal skin ERK1/2 UACC257 COLO829 MALME3M SK-MEL-2 451Lu Pt2 Pt3 Pt4 Pt5 Pt6 Pt7 Pt8 Pt9 Pt10 Pt11 Pt12 HEMd HEMI HEMI HEMd 50 *

* ACTIN

PM Melanoma Patient tumors cell lines e WA09 -C6 (Control) WA09 -C3 (NGLY1-KO)

d WA09 hESCs NGLY1-CRISPR Clone 6 (C6) Clone 3 (C3)

WA09 hESCs NGLY1-CRISPR WA09 C6 C3 C4 (KDa) 75 NGLY1 50 NANOG

50 POU5F1

50 ACTIN

f NGLY1Pt1i-509 hiPSCs

g NGLY1Pt1i-509 (NGLY1-deficient)

hiPSCs Con Pt1 (KDa) 75 NGLY1

50 NANOG

50 POU5F1

50 ACTIN

Cytokine Profiling and Neutralisation Specific antibodies against human INFβ1 and IL-29 (R&D Systems, U-PLEX Human Interferon Combo assay kits and a SECTOR Imager Minneapolis, MN; Supplementary Table S2) were used to 2400 (Meso Scale Discovery, Rockville, MD) were used to measure neutralise the cytokines in cell samples, while the IgG isotype cytokine contents in conditioned medium samples of cells with (Jackson ImmunoResearch Laboratories, West Grove, PA) was indicated treatment, according to the manufacturer’s instructions. applied to control samples. Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1541 Fig. 1 NGLY1 expression in normal and melanoma cells. (a) The expression of the NGLY1 gene in cells at the transcriptional level was measured using qRT-PCR. Orange shading: primary melanocytes (PM). Blue shading: human melanoma cell lines. Pink shading: tumour samples of melanoma patients. Blue dot: undetectable NGLY1 transcript in the sample. Red asterisk: cell samples collected at high passage numbers ( > 16). All data were presented as mean ± standard deviation (n = 3; *P < 0.05, Mann–Whitney U test). The expression level of the ACTB gene in each sample was used as internal control for normalization. Gene expression levels in HEMl cells were used as comparison standards to calculate relative expression values. (b) The protein levels of NGLY1 detected using western blotting in cell samples. Blue shading: human normal cells. Yellow shading: human melanoma cell lines. (c) Representative images for immunohistochemistry staining of NGLY1 in normal skin and melanoma tumour tissues. With the same staining condition, the expression of NGLY1 was positively stained in the melanoma tumour tissue but not detected in the normal skin tissue. (d) WA09 hESC clones with and without the gene editing-mediated ablation of NGLY1 expression. Upper panel: cell morphology. Lower panel: the expression NGLY1 and pluripotency markers NANOG and POU5F1 detected by western blotting in the cells. WA09: parental WA09 hESCs. WA09-C6: a cell clone of WA09 hESCs derived from a gene-editing and selection process without acquiring disruptive mutations in the NGLY1 gene. WA09-C3 and WA09-C4: two NGLY1-deficent cell clones of WA09 hESCs independently derived from a gene-editing and selection process. (e) The positive staining of pluripotency markers in WA09-C6 and WA09-C3 hESCs. (f) NGLY1Pt1i-509 hiPSCs established from cell reprogramming in NGLY1-deficient patient-derived dermal fibroblasts. Upper panel: cell morphology. Lower panel: the expression NGLY1 and pluripotency markers NANOG and POU5F1 detected by western blotting in the cells. (g) The positive staining of pluripotency markers in NGLY1Pt1i-509 hiPSCs

In vivo Studies RESULTS The animal work in this study was completed using an animal NGLY1 is highly upregulated in melanoma cells but dispensable study service provided by the translational core laboratory at the for the vitality and pluripotency of pluripotent stem cells University of Maryland, Baltimore. All experimental procedures In contrast to human normal melanocytes, a majority of tested and protocols utilizing mice were approved by the Institutional melanoma cell lines showed upregulation of NGLY1 (Fig. 1a, b). Animal Care and Use Committee at the University of Maryland. The significant upregulation of NGLY1 was also observed in The procedures were provided as part of Supplementary Materials patients’ tumour samples (Fig. 1a, c; Supplementary Table S3). and Methods. Human PSCs, including undifferentiated hESCs and induced pluripotent stem cells (hiPSCs), are considered as normal cells in Proteomics Analysis a unique cellular state, while similar to cancer cells regarding The detailed procedures were provided as part of Supplementary certain features.16 Normal melanocytes (HEMl)-derived hiPSCs, Materials and Methods. compared with HEMl cells, showed a similar expression level of NGLY1, much lower than that observed in melanoma cells (Fig. 1b). Chemical Synthesis and Characterisation of NGLY1 Inhibitors These findings suggest a potentially critical function of NGLY1 for The detailed procedures of chemical synthesis and characterisa- cancer cells to sustain their viability or oncogenic signalling. tion for the novel analogs of N-acetylglucosamine (GlcNAc)- Through gene editing, we obtained NGLY1-knockout (NGLY1-KO) linked asparagine were provided as part of Supplementary clones and an editing-escaping (control) clone of WA09 hESCs. Materials and Methods. Z-VAD-fmk were purchased from Despite the deficiency of NGLY1 observed in WA09-C3 and WA09- Millipore Sigma (St. Louis, MO). WRR139 was synthesised and C4 hESCs (Fig. 1d), their morphology, viability and pluripotency characterised according to the chemical approaches previously were highly comparable to that of WA09-C6 and parental WA09 described.15 hESCs with NGLY1 expression (Fig. 1d, e; Supplementary Figure S1A and S1B). We also reprogrammed the NGLY1- Computational Modeling deficient patient’s dermal fibroblasts into hiPSCs (Fig. 1f). Like The procedures were provided as part of Supplementary Materials the NGLY1-KO hESCs, the patient-derived hiPSCs without NGLY1 and Methods. expression can be continuously cultured and maintain typical hPSC morphology, molecular features and the capacity of forming Production of recombinant human NGLY1 and RNase B embryoid bodies (EBs) containing differentiated cells that are deglycosylation assay associated with three germ layers (Fig. 1g; Supplementary The procedures for generating recombinant human NGLY1 and Figure S1C). Our results reveal that, while highly upregulated in testing its enzymatic activity were provided as part of Supplemen- melanoma cells, NGLY1 appears to be dispensable for the vitality tary Materials and Methods. of human normal cells even in a highly sensitive state like the embryonic stage. Statistical Analysis The significance of differences in comparisons was primarily NGLY1 suppression disturbs proteasome-mediated protein determined by the two-tailed Student’s t-test for a two-group degradation and induces stress response signalling-associated comparison, unless stated otherwise in the figure legends. The apoptosis in melanoma cells association of NGLY1 staining results and pathological condi- Two independent NGLY1-targeting shRNA sequences (NGLY1- tions in normal skin and melanoma tumour tissues was shRNA645 and NGLY1-shRNA647) and a non-targeting shRNA (NT- examined using a 2 × 2 contingency with the two-tailed Fisher’s shRNA) sequence were cloned into doxycycline (dox)-inducible, exact test. polycistronic green fluorescence protein (GFP)-shRNA expression constructs (TransOMIC Technologies, Huntsville, AL; Fig. 2a). We Data and materials availability generated the stable clones with the shRNA expression constructs The gene expression array data have been deposited with a link to in MALME3M, UACC257, SK-MEL-2, and COLO829 melanoma cells. an accession number GSE106936 in the Gene Expression Omnibus The stable clones of melanoma cells showed clear GFP expression (GEO). Other data included within the article to support the upon dox treatment (Fig. 2b), indicating the expression of shRNA findings of this study are available from the corresponding author sequences. Compared with the cells expressing the NT-shRNA, the upon reasonable request. The biological samples and novel expression of NGLY1 was largely suppressed by NGLY1-shRNA645 compounds used in this study may be distributed upon request and NGLY1-shRNA647 and hardly detectable in the stable clones and under institutional material transferring agreements or a of MALME3M, UACC257 and SK-MEL-2 cells with 48-hour licensing process. treatment of dox (Fig. 2c). Upon NGLY1 knockdown, melanoma Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1542

a shRNA sequences NGLY1-shRNA645: 5’ 3’ 3’ pZIP-TRE3GS 5′mir 3′mir ZsGreen TRE3GS PGK Tet-On 3G TA SV40 PuroR NGLY1-shRNA647: 5’ shRNA loop

bcMALME3M UACC257 SK-MEL-2 dMALME3M SK-MEL-2 MALME3M SK-MEL-2 ++–– –– ++–– –– NT-shRNA (2 μM Dox, 36 h) shRNA shRNA shRNA – + – – + – – + – – + – NGLY1-shRNA (2 μM Dox, 36 h) Without Dox With Dox With Dox ––+ – – + ––+ – – + Bortezomib (0.5 μM, 36 h) (KDa) Tun NT 647 645 Tun NT 647 645 Tun NT 647 645 ++++++ ++++++ Cycloheximide (40 μg/mL, 36 h) 75 NGLY1 250 (kDa) 150 50 ATF4 100 75 25 GADD153 50 100 GRP94 UACC257_NT-shRNA KDEL 37 75 GRP78 25 20 50 15 ACTIN 10 Amido black staining IB: Ubiquitin e HDF51 MALME3M UACC257 SK-MEL-2 f 100 HDF51 (NT) MALME3M (NT) 6 6 6 6 10 10 10 10 HDF51 (shRNA645) MALME3M (shRNA645)

5 5 5 5 UACC257 (NT) SK-MEL-2 (NT) 10 10 10 10 80 UACC257 (shRNA645) SK-MEL-2 (shRNA645) 104 104 104 104

3 3 3 3 60 * NT-shRNA 10 10 10 10 Annexin V-AF555 Annexin V-AF555 Annexin V-AF555 Annexin V-AF555 Annexin 2 2 2 2 * 10 10 10 10 * 2 3 4 5 6 2 3 4 5 6 2 3 4 5 6 2 3 4 5 6 UACC257_NGLY1-shRNA645 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 40 GFP GFP GFP GFP 106 106 106 106

105 105 105 105 20 Annexin V-positive cells (%) V-positive Annexin 104 104 104 104 0 103 103 103 103 Annexin V-AF555 Annexin V-AF555 Annexin V-AF555 Annexin V-AF555 Annexin 2 2 2 2

NGLY1-shRNA645 10 10 10 10 102 103 104 105 106 102 103 104 105 106 102 103 104 105 106 102 103 104 105 106 GFP GFP GFP GFP HDF51 (NT) g MALME3M (NT) UACC257 (NT) SK-MEL-2 (NT) Without doxycycline treatment HDF51 (shRNA645) MALME3M (shRNA645)UACC257 (shRNA645)SK-MEL-2 (shRNA645) MALME3M (72 h) MALME3M (72 h) MALME3M (72 h) MALME3M (72 h)

120 NT-shRNA 120 120 120 NT-shRNA NGLY1-shRNA645 NGLY1-shRNA645 NGLY1-shRNA647 100 100 100 NGLY1-shRNA647 100

80 80 80 80

60 60 60 60

40 40 40 40 NT-shRNA NT-shRNA 20 20 NGLY1-shRNA645 20 20 NGLY1-shRNA645 * * Cell viability (% of control) Cell viability (% of control) NGLY1-shRNA647 Cell viability (% of control) Cell viability (% of control) NGLY1-shRNA647 0 0 0 0 1101001000 1 10 100 10000.010.1 1 10 100 0.1 1 10 Cisplatin (μM) Dacarbazine (μM) Vemurafinib (μM) Doxycycline (μM)

SK-MEL-2 (72 h) SK-MEL-2 (72 h) SK-MEL-2 (72 h) SK-MEL-2 (72 h) 120 120 120 120 NT-shRNA NGLY1-shRNA645 100 NGLY1-shRNA647 100 100 100

80 80 80 80

60 60 60 60

40 40 40 40 NT-shRNA NT-shRNA NT-shRNA 20 20 NGLY1-shRNA645 20 NGLY1-shRNA645 20 NGLY1-shRNA645 *

Cell viability (% of control) Cell viability (% of control) Cell viability (% of control) NGLY1-shRNA647 Cell viability (% of control) * NGLY1-shRNA647 NGLY1-shRNA647 0 0 0 0 1 10 100 1000 1 10 100 10000.01 0.1 1 10 100 0.1 1 10 100 Cisplatin (μM) Dacarbazine (μM) Vemurafinib (μM) Doxycycline (μM)

h MALME3M_NT-shRNA (72 h) MALME3M_shRNA645 (72 h) MALME3M_shRNA647 (72 h) MALME3M (72 h)

120 120 Doxycycline 120 Doxycycline 5 NT-shRNA Dacarbazine Dacarbazine 4 NGLY1-shRNA645 Combination (Dox:Daca = 1:20) Combination (Dox:Daca = 1:20) 100 100 100 3 NGLY1-shRNA647 2 1 80 80 80 0.4 60 60 60

40 40 40 0.2 (Dox with Daca) (Dox

Doxycycline Combination index 20 Dacarbazine 20 20 Cell viability (% of control) Combination (Dox:Daca = 1:20) Cell viability (% of control) Cell viability (% of control) 0 0 0 0.0 0.1 1 10 100 1000 0.1 1 10 100 1000 0.1 110100 1000 0.0 0.2 0.4 0.6 0.8 1.0 Dose (μM) Dose (μM) Dose (μM) Fractional affected

SK-MEL-2_NT-shRNA (72 h) SK-MEL-2_shRNA645 (72 h) SK-MEL-2_shRNA647 (72 h) SK-MEL-2 (72 h)

120 Doxycycline 120 Doxycycline 120 Doxycycline 2 NT-shRNA Dacarbazine Dacarbazine Dacarbazine NGLY1-shRNA645 Combination (Dox:Daca = 1:15) 100 Combination (Dox:Daca = 1:15) 100 100 Combination (Dox:Daca = 1:15) NGLY1-shRNA647

80 80 80

60 60 60 1

40 40 40 (Dox with Daca) (Dox 20 20 20 Combination index Cell viability (% of control) Cell viability (% of control) Cell viability (% of control) 0 0 0 0 0.1 1 10 100 1000 0.1 1 10 100 1000 0.1 1 10 100 1000 0.0 0.2 0.4 0.6 0.8 1.0 Dose (μM) Dose (μM) Dose (μM) Fractional affected Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1543 Fig. 2 ER stress-associated apoptosis and synergistic anticancer responses induced by NGLY1 knockdown in melanoma cells. (a) The doxycycline (dox)-inducible pZIP-TRE3GS expression vector of non-targeting shRNA and NGLY1-targeting shRNA sequences. (b) The stable clones of UACC257 cells with dox-inducible shRNA. Cells with induced NGLY1-shRNA645 (green cells) showed morphological features of apoptosis, including shrinkage and fragmentation. Cells with induced non-targeting (NT)-shRNA maintain a morphology similar to the cells before dox induction. Cells were imaged after the treatment of 2 µM dox for 72 h. BF: bright field. ZsGreen: green fluorescence protein. (c) ATF4 and GADD153 signalling was activated by the shRNA-mediated knockdown of NGLY1 in melanoma cells. Tun: 2 µM tunicamycin for 24 h. (D) The accumulation of ubiquitinated proteins was detected using western blotting in MALME3M and SK-MEL-2 cells with NGLY1 knockdown. (e) The representative quadrant plots of flow cytometry analysis to detect apoptosis in normal (HDF51) and melanoma cells with NGLY1 knockdown. The cell samples were collected for analysis after the 72-hour induction of shRNA expression. (f) The quantitative results of flow cytometry analysis to detect apoptosis. All data were presented as mean ± standard deviation (n = 3; *P < 0.05, t-test) in the bar graph. (g) The dose-dependent suppression of viability in MALME3M and SK-MEL-2 cells with the indicated dox-inducible shRNA in response to cisplatin, dacarbazine, vemurafenib and dox treatment. All data were presented as mean ± standard deviation (n = 3, *P < 0.05, logistic regression). (h) The synergistic anticancer responses of NGLY1 knockdown and dacarbazine treatment for 72 h in MALME3M and SK-MEL-2 cells. The cell viability curves of combinatorial treatment were plotted according to the doses of dox used in the treatment. Combination indexes were calculated using Calcusyn software. A combination index value < 1 was considered synergistic. A combination index value < 0.2 was considered highly synergistic. All cell viability data were presented as mean ± standard deviation (n = 3)

cells showed morphological features of apoptosis, including Proteomic analysis reveals unique peptide signatures in shrinkage, fragmentation and detachment (Fig. 2b). melanoma cells in response to NGLY1 suppression As shown in Fig. 2c, the NGLY1 knockdown-induced upregula- Upon NGLY1 suppression, ENGase has better access to the tion of ATF4 and GADD153 was detected in melanoma cells. In substrates and generates more products of glycopeptides addition, NGLY1 knockdown hindered proteasome-mediated containing GlcNAc-asparagine residues in cells (Supplementary protein degradation, indicated by the accumulation of ubiquiti- Figure S5A).1 Using LC-MS/MS-based proteomics analysis, we have nated proteins in cycloheximide-treated MALME3M and SK-MEL-2 identified peptides containing GlcNAc-asparagine residues cells (Fig. 2d). This finding is consistent with the previously denoted as N(HexNAc) in both control and NGLY1-knockdown observed suppression of ERAD in cells with NGLY1 malfunc- samples (Supplementary Figure S5B and S5C). Although the tion.1,4,5,15,17–19 Since GADD153 is an important mediator for ER quantities of peptides containing GlcNAc-asparagine that can be stress-associated apoptosis,13 our findings suggest that ER stress detected among the biological replicates of different cell samples signalling-mediated apoptosis may contribute to the death of appeared to vary, compared with control cells, melanoma cells melanoma cells with NGLY1 suppression. Using flow cytometry with NGLY1 knockdown reproducibly showed higher contents of analysis, a substantial increase of apoptosis was detected in peptides containing GlcNAc-asparagine residues, indicating the NGLY1-knockdown melanoma cells, which was absent in the functional defect of NGLY1 in the cells (Supplementary Fig- cancer cells expressing NT-shRNA and normal cells expressing ure S5D). Many proteins also presented differential abundance in NGLY1-targeting shRNA (Fig. 2e, f). Overexpression of exogenous melanoma cells in response to NGLY1 knockdown (Supplementary human NGLY1 and knockdown of GADD153 both attenuated Table S4). Among the proteins with reduced abundance in the apoptosis induced by NGLY1 knockdown in SK-MEL-2 and NGLY1-knockdown cells, several of them (e.g., VCP, PDIA4, HSPA5 UACC257 cells (Supplementary Figure S2). Taken together, stress and HIST1H4A) have been linked to the survival and drug response-associated, GADD153-mediated apoptosis contributes to resistance of cancer cells.22–25 Thus, part of the anti-melanoma NGLY1 knockdown-induced melanoma cell death. responses associated with NGLY1 inhibition may be attributed to the modulation of these gene products. NGLY1 suppression sensitises melanoma cells to the treatment of DNA alkylating agents Global gene expression profiling uncovers pleiotropic effects of Unlike the treatment of an ER stress inducer, tunicamycin, NGLY1 suppression on melanoma cells NGLY1 knockdown activates the transcription factor ATF4 and Using transcriptomic analysis, we identified a group of genes its downstream apoptotic factor GADD153 without upregulating (~750 gene probes corresponding to ~700 genes) that were ER chaperones GRP78/94 (Fig. 2c). Many chemotherapeutic significantly (P < 0.01) and commonly upregulated or down- drugs including DNA alkylating agents also induce GADD153 in regulated between control and NGLY1-knockdown melanoma cancer cells.20 Thus, NGLY1 suppression may synergise with DNA cells. The hierarchical clustering of all the cell samples based on alkylating agents like dacarbazine and temozolomide to the expression of these genes showed that, within the same cell eliminate melanoma cells, at least partially, through intensified line, all the NGLY1-kockdown samples were similar and segre- activation of GADD153. We tested whether NGLY1 suppression gated from the control samples (Fig. 3a). Data analysis with an enhances the anticancer activity of dacarbazine and temozolo- additional filtering criterion (expression fold change ≥ 2) showed mide that are commonly used to treat melanoma. Unlike that NGLY1 suppression appeared to primarily induce gene MALME3M cells with the BRAFV600E mutation and high upregulation. Many of these upregulated genes, including the sensitivity to vemurafenib,21 SK-MEL-2 cells with the NRASQ61R IFNβ1 and IL-29 genes, are highly associated with cytokine mutation21 are resistant to vemurafenib (Fig. 2g). The knock- responses in cells (Fig. 3b). The expression of differentially down of NGLY1 compromised the viability of MALME3M and SK- expressed genes was also validated using qRT-PCR (Fig. 3c). In MEL-2 cells (Fig. 2g) in viability assays where we also observed a addition to the cytokine signalling-relevant genes, many genes highly synergistic effect of NGLY1 knockdown in combination like XAF, ATF3, PMAIP1 (NOXA), AXUD1 and CDKN2C that have been with the cytotoxicity of either dacarbazine or temozolomide liked to anticancer activity26–31 were significantly upregulated in (Fig. 2h; Supplementary Figure S3 and S4). These results indicate the NGLY1-knockdown cells, while genes like FABP7, CRYAB and that the suppression of NGLY1 could overcome melanoma cells GAPDHS that have been associated with the survival, proliferation with resistance to BRAF inhibitors as well as sensitise the cells to and invasiveness of cancer cells or with a poor prognosis conventional chemotherapy agents that frequently lead to in melanoma patients32–34 were significantly downregulated unsatisfactory outcomes in the treatment of patients with (Supplementary Table S5). Ontology analysis showed that melanoma. ~60 differentially expressed genes (P < 0.01, expression fold Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1544 a b

Gene probes (expression difference: P >0.01 or Fold change <2) 3 Gene probes (expression difference: P <0.01 & Fold change ≥2) 2

1 RSAD2 NGLY1 0 6 –1 HLA-F –2

~750 gene probes IFIT3 –3 OASL 4 -value HLA-B p SK-MEL-2_shRNA647 (2) SK-MEL-2_shRNA645 (2) SK-MEL-2_shRNA647 (1) SK-MEL-2_shRNA645 (1) MALME3M_shRNA647 (1) MALME3M_shRNA647 (2) MALME3M_shRNA645 (1) MALME3M_shRNA645 (2) UACC257_shRNA647 (1) UACC257_shRNA645 (1) UACC257_shRNA647 (2) UACC257_shRNA645 (2) MALME3M_NT-shRNA (1) MALME3M_NT-shRNA (2) UACC257_NT-shRNA (1) UACC257_NT-shRNA (2) COLO829_NT-shRNA (1) COLO829_NT-shRNA (2) SK-MEL-2_NT-shRNA (1) SK-MEL-2_NT-shRNA (2)

COLO829_shRNA647 (2) COLO829_shRNA647 (1) COLO829_shRNA645 (1) COLO829_shRNA645 (1) CCL5 FABP7 IFIT2 HCP5 ATF3

–Log10 IFNB1 IFITM1 2 CDCA7

IL29

–2 –1 0 1 2 3 Log2 fold change

c UACC257_NT-shRNA SK-MEL-2_NT-shRNA MALME3M_NT-shRNA COLO829_NT-shRNA UACC257_NGLY1-shRNA645 SK-MEL-2_NGLY1-shRNA645 MALME3M_NGLY1-shRNA645 COLO829_NGLY1-shRNA645

26 FABP7 RSAD2 CCL5 * IFN1

4 * 2 * * * 2 * * 2 * * ** * * 20

2–2 * 2–4 Relative normalized expression Relative * *

d Response to stimulus: Biologocal adhesion Behavior (GO:0007610) (GO:0022610) Immune response (GO:0006955) Cellular component organization Response to biotic stimulus (GO:0009607) or biogenesis (GO:0071840) Response to external stimulus (GO:0009605) Response to stress (GO:0009650) Response to stimulus (GO:0050896) (GO:0065007) Metabolic process: Biological regulation Cellular process Biosynthetic process (GO:0009058) (GO:0009987) Catabolic process (GO:0009056) Immune system process Coenzyme metabolic process (GO:0006732)

(GO:0002376) Nitrogen compound metabolic process (GO:0006807) Multicellular organismal Primary metabolic process (GO:0044238) process (GO:0032501) Metabolic process (GO:0008152) Developmental process (GO:0032502) Cellular process:

Cell communication (GO:0007154) Cell cycle (GO:0007049) Localization Locomotion (GO:0051179) Cell proliferation (GO:0008283) (GO:0040011) Cellular component movement (GO:0006928)

e Gene probes (Expression diffrence: P >0.01 or Fold change <2) f Gene probes (Expression diffrence: P >0.01 or Fold change <2) Gene probes (Expression diffrence: P < 0.01 & Fold change ≥2) Gene probes (Expression diffrence: P < 0.01 & Fold change ≥2) 5 4

4 3

3 -value -value p p 2 2 –Log10 –Log10 1 1

0 0 –2 –1 0 1 2 –2 –1 0 12 Log2 fold change Log2 fold change

change ≥ 2) are highly involved in multiple biological processes melanoma transcriptomes, the disruption of NGLY1 expression in (Fig. 3d), including immune and stress responses, primary normal hPSCs and their differentiated derivatives caused limited metabolic process, cell communication, and cell cycle. In contrast changes in their gene expression networks (Fig. 3e, f). None of to substantial perturbation induced by NGLY1 knockdown in the differentially expressed genes (P < 0.01, expression fold Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1545 Fig. 3 Differential gene expression caused by NGLY1 suppression in melanoma cells, hESCs and the differentiated derivatives of hESCs. SK- MEL-2, COLO829, UACC257 and MALME3M melanoma cells with the expression of the indicated inducible shRNA due to the treatment of 2 µM dox for 48 h were collected for RNA isolation and global gene expression profiling. WA09, WA09-C6, WA09-C3, WA09-C4 hESCs and their differentiated derivatives were also collected for analysis. Samples of two biological replicates for each setting were analysed. (a) A heat map representation of ~750 probes that measured the relative expression levels of differentially expressed genes (P < 0.01, t-test between control and knockdown cells) in melanoma cell samples expressing the indicated shRNA. Red dots: melanoma cells with NGLY1 knockdown. Green dots: control cells. (b) Selected genes that were differentially expressed (P < 0.01 and fold change ≥ 2) in the control and NGLY1-knockdown melanoma cells were annotated in a volcano plot of fold change vs. significance. (c) The qRT-PCR validation of selected genes that were differentially expressed in the control and NGLY1-knockdown melanoma cells (n = 3, *P < 0.05, t-test). The expression level of the ACTB gene in each sample was used as internal control for normalisation. Gene expression levels in SK-MEL-2 cells with NT-shRNA were used as comparison standards to calculate relative expression values. (d) Gene ontology analysis revealed that genes differentially expressed (P < 0.01 and fold change ≥ 2) due to NGLY1 suppression in melanoma cells were highly enriched in multiple biological processes. (e) A volcano plot of fold change vs. significance for selected genes that were differentially expressed (P < 0.01 and fold change ≥ 2) in control and NGLY1-deficient WA09 hESCs. (f) A volcano plot of fold change vs. significance for selected genes that were differentially expressed (P < 0.01 and fold change ≥ 2) in the embryoid bodies of control and NGLY1-deficient WA09 hESCs gone through 6 days of non-directed differentiation

a 72-h Dox treatment 500 IFNβ 400 IL-29 * b 300 shRNA645 (24-h Dox) 200 500 * IFNβ 90 400 IL-29 70 * 300 50 * 200 30 100 *

Concentration (pg/ml) Concentration 10 100

Concentration (pg/ml) Concentration 40 * 2 20 * * 0 0

UACC257 (Control) SK-MEL-2 (Control) HDF51_NT-shRNA HDF51_NT-shRNA HDF51_shRNA645 HDF51_shRNA645 UACC257 (NGLY1 OE)SK-MEL-2 (NGLY1 OE) UACC257_NT-shRNAUACC257_shRNA645SK-MEL-2_NT-shRNA UACC257_NT-shRNAUACC257_shRNA645SK-MEL-2_NT-shRNA MALME3M_NT-shRNAMALME3M_shRNA645 SK-MEL-2_shRNA645MALME3M_NT-shRNAMALME3M_shRNA645 SK-MEL-2_shRNA645

c 160 SK-MEL-2_NGLY1-shRNA645 (IgG isotype) 160 SK-MEL-2_NGLY1-shRNA645 (IgG isotype) SK-MEL-2_NGLY1-shRNA645 (anti-IFNβ1 IgG) SK-MEL-2_NGLY1-shRNA645 (anti-IL29 IgG) MALME3M_NGLY1-shRNA645 (IgG isotype) MALME3M_NGLY1-shRNA645 (IgG isotype) MALME3M_NGLY1-shRNA645 (anti-IFNβ1 IgG) MALME3M_NGLY1-shRNA645 (anti-IL29 IgG) 100 100 d SK-MEL-2 MALME3M * * * * NT 645 NT 645 (shRNA) 80 80 (KDa) p-TBK1 (Ser172) 75 * * * TBK1 60 60 75 50 p-IRF3 (Ser396)

40 40 50 IRF3

IRF7 50 Cell viability (% of mock treatment) Cell viability (% of mock Cell viability (% of mock treatment) Cell viability (% of mock 20 20 50 Actin

0 0 0 0.25 12 0 0.25 12 Concentration of IgG (μg/ml) Concentration of IgG (μg/ml) Fig. 4 NGLY1 suppression enhanced the production of IFNβ1 and IL-29 that contributes to viability reduction in melanoma cells. (a) The contents of IFNβ1 and IL-29 in the conditioned media of UACC257 and SK-MEL-2 cell clones with the indicated treatment were measured by cytokine profiling. (b) The NGLY1 knockdown-induced upregulation of IFNβ1 and IL-29 was significantly attenuated by the overexpression of exogenous human NGLY1 in the cells. (c) Left panel: the attenuation of NGLY1 knockdown-induced viability reduction by the treatment of specific IFNβ1 neutralisation antibody in the cells. Right panel: the attenuation of NGLY1 knockdown-induced viability reduction by the treatment of specific IL-29 neutralisation antibody in the cells. NGLY knockdown was induced by the treatment of 2 µM dox for 72 h in the cells. (d) The enhanced expression and activation of IRF3, IRF7 and their upstream kinase TBK1 was detected in SK-MEL-2 and MALME3M cells with NGLY1 knockdown. The serine phosphorylation of IRF3 and TBK1 indicates their activity. NT: non-targeting shRNA. 645: NGLY1-targeting shRNA645. All data were presented as mean ± standard deviation (n = 3; *P < 0.05, t-test) Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1546

a Melanoma cells with Injection of 1×10^6 cells dox-inducible NT-shRNA Tumor size: ~150 mm^3 Cells collected Dosing with dox for injection in drinking water Monitoring the health condition and tumor growth in mice for ~5 weeks

Melanoma cells with Injection of 1×10^6 cells Tumor size: ~150 mm^3 dox-inducible NGLY1-shRNA645 Cells collected Dosing with dox for injection in drinking water Monitoring the health condition and tumor growth in mice for ~5 weeks

bcAverage tumor volume (NT-shRNA): 1999.78 ± 966.32 mm^3 Average tumor volume (shRNA645): 1000.09 ± 864.38 mm^3 SK-MEL-2 tumors 4000 NT-shRNA 5R NT-shRNA shRNA645 Tumor-2R NGLY1-shRNA645 1600 Tumor-2L 3L 3R4L 4R 5L 5R 2L 2R 4L 4R (Factorial ANOVA; F = 8.537, P < 0.01) 1200 Tumor-4R (kDa) 800 75 NGLY1 3000 3L 5L 400

Tumor volume (mm^3) volume Tumor 4L 0 25 0714 21 28 35 GADD153 Days (Dox treatment) 2000 4R 50 IRF3 4R 3R ZsGreen Tumor volume (mm^3) volume Tumor 1000 4L 25 2L Actin 37 2R 0 0 7 14 21 28 34 0 7 14 21 28 34 Days (Dox treatment)

d NT-shRNA (Tumor 4R) NGLY1-shRNA645 (Tumor 2L)

Fig. 5 The in vivo antitumour activity of targeting NGLY1 in melanoma cells. (a) A schematic illustration of animal study design to test the in vivo antitumour efficacy of NGLY1 suppression in melanoma. (b) The volume changes of xenografted SK-MEL-2 tumours with the induction of NT-shRNA (n = 10) and NGLY1-shRNA645 (n = 8) for 35 days. NGLY1 knockdown was a significant factor that affected the tumour volume (Factorial ANOVA; F = 8.537, P < 0.01). Tumours were harvested at the end of the study for western blotting analysis. Bars: median tumour volumes at the indicated time points. Inset: the volume changes of three tumours with NGLY1-targeting shRNA that initially increased their size but showed regression at the end of the study. (c) The expression of NGLY1, GADD153, IRF3, and GFP (ZsGreen) proteins in selected tumours was analysed by western blotting. (d) The enhanced expression of IL-29 in the tumour tissues with NGLY1 knockdown was detected by immunofluorescence staining

change ≥ 2) identified in NGLY1-knockdown melanoma cells NGLY1 significantly attenuated the NGLY1 knockdown-mediated showed significant expression alterations in NGLY1-deficient induction of IFNβ and IL-29 in melanoma cells (Fig. 4b). Using hPSCs and their differentiated derivatives, highlighting the IFNβ1- and IL-29- neutralisation antibodies but not their IgG fundamental differences of normal and malignant cells in isotypes, we partially rescued the NGLY1 knockdown-triggered response to NGLY1 inhibition. inhibition of viability in SK-MEL-2 and MALME3M cells (Fig. 4c). In addition, a considerable overlap was found between the NGLY1 suppression induces cytokine surges that contribute to differentially expressed genes in melanoma cells with IL-29 melanoma cell death treatment35 and NGLY1 knockdown (Supplementary Table S5). In contrast to the conditioned medium of control cells, the Notably, the anti-melanoma activity of IFNβ and IL-29 has been conditioned medium of NGLY1-knockdown melanoma cells demonstrated in previous studies.35–37 These findings definitively contained a significantly higher amount of IFNβ and/or IL-29 illustrate that NGLY1 suppression-induced cytokine surges and (Fig. 4a), consistent with our findings from gene expression relevant cellular responses contribute to anticancer effects. Unlike profiling. The overexpression of exogenous human melanoma cells, human normal fibroblasts showed no significant Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1547 a Carbons of the co-crystallographic ligand in mouse NGLY1 Carbons of the ligand docked in human NGLY1 homology model (–6.3 kcal/mol)

Docking

Cys309 Human NGLY1 homology model

O O OMe H N O N N F Human NGLY1 H H O O Z-VAD-fmk b NM-322 (–6.9 kcal/mol) NM-348 (–6.3 kcal/mol) NM-350 (–6.8 kcal/mol) NM-354 (–6.8 kcal/mol)

Compound R R O O HN N NM-322 H O O O O NM-348 S O NH O O O CYS309 NM-350 CYS309 CYS309 O N CI CYS309 H O O O NM-354 N O

c 200 μM (2h) d NM-322 NM-348

(kDa) RNase B ZVAD WRR NM NM NM NM 140 120 Veh 348 M 250 only fmk 139 322 350 354 120 150 100 100 100 * 75 80 * 80 * 50 * 60 37 60 HDF51 40 HDF51 Normal fibroblasts 40 HDF418 UACC257 from two individuals 25 20 UACC257 20 SK-MEL-2 SK-MEL-2

20 Cell viability (% of control) Cell viability (% of control) 0 0 RNase B (g) 1 10 100 1000 1 10 100 1000 15 RNase B (dg) Concentration (μM) Concentration (μM)

10 NM-350 NM-354 Vehicle/ NM-350/ Z-VAD-fmk/ 120 120 Bortezomib Bortezomib Bortezomib 100 100 (kDa) – + – + – + PNGase F * * 150 80 80 *

100 NFE2L1 60 60 75 40 HDF51 40 HDF51 UACC257 UACC257 20 SK-MEL-2 20 SK-MEL-2 Actin Cell viability (% of control) 0 Cell viability (% of control) 0 1 10 100 1000 1 10 100 1000 Concentration (μM) Concentration (μM) efSK-MEL-2 (72 h) SK-MEL-2 (72 h) SK-MEL-2 (48 h) SK-MEL-2 (48 h)

140NM-322 4 140 NM-322 4 Dacarbazine Dacarbazine 120 Combination (Daca:NM-322=1:1.25) 120 Combination (Daca:NM-322=1:1.25) 100 3 100 3 80 80 2 2 60 60

40 1 40 1 Combination index Combination index 20 (NM-322 with Daca) 20 (NM-322 with Daca) Cell viability (% of control) Cell viability (% of control) 0 0 0 0 1 10 100 10000.0 0.2 0.4 0.6 0.8 1.0 0.0001 0.01 1 100 0.0 0.2 0.4 0.6 0.8 1.0 Dose (μM) Fractional affected Dose (μM) Fractional affected Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1548 Fig. 6 Anticancer responses induced by novel covalent modifiers that target the catalytic site of human NGLY1 in melanoma cells. The computational homology model of human NGLY1 core domain was generated and used for studying interactions between NGLY1 and novel small molecules that are designed to covalently modified and inactivate the catalytic site of NGLY1. (a) The most favourable binding pose of Z- VAD-fmk, a short peptide with NGLY1 and caspase inhibitory activity, in the human NGLY1 homology model superimposed to the conformation of Z-VAD-fmk bound to mouse NGLY1 in a co-crystalised structure. (b) Novel small molecules (NM-322, NM-348, NM-350, and NM-354) that mimic a GlcNAc-conjugated asparagine in the NGLY1 substrates of NGLY1 and contain strategically positioned electrophilic groups bound to the human NGLY1 homology model in computational docking and showed their high binding affinities with the electrophilic groups pointed towards Cys309 in close proximity at the human NGLY1 catalytic site. (c) Upper panel: the 2-hour reaction of covalent modifiers, including Z-VAD-fmk (20 µM), WRR139 (5 µM), NM-322, NM-348, NM-350 and NM-354, with human NGLY1 suppressed its activity in the deglycosylation of denatured RNase B. Blue arrowhead: recombinant NGLY1-FLAG. RNase B (g): glycosylated RNase B. RNase B (dg): deglycosylated RNase B. Veh: vehicle (DMSO) treatment. M: molecular weight marker. Lower panel: the deglycosylation of NFE2L1 altered by the treatment of 20 µM Z-VAD-fmk and 200 µM NM-350 in bortezomib-treated HEK293T cells. The cells were pretreated with vehicle (DMSO), Z-VAD-fmk and NM-350 for 24 h and subsequently subjected to concomitant treatment with 10 µM bortezomib for an additional 16 h. Cell lysates reacted with and without 500 units of PNGase F for 2 h were analysed using western blotting. Red arrowhead: fully glycosylated NFE2L1. Orange arrowhead: partially glycosylated NFE2L1. Green arrowhead: deglycosylated and truncated NFE2L1. (d) The dose-dependent suppression of cell viability was preferentially induced by the novel NGLY1 inhibitors in melanoma cells compared with normal cells (*P < 0.05, logistic regression). (e) The synergistic effect was observed between NM-322 and dacarbazine in the suppression of melanoma cell viability. The cell viability curve of combinatorial treatment was plotted according to the doses of dacarbazine used in the treatment. (f) The synergistic effect was observed between NM-350 and bortezomib in the suppression of melanoma cell viability. The cell viability curve of combinatorial treatment was plotted according to the doses of bortezomib used in the treatment. All the data of cell viability tests were presented as mean ± standard deviation (n = 3)

changes in the production and release of IFNβ and IL-29 in molecules that may inactivate human NGLY1. Since Z-VAD-fmk (a response to NGLY1 knockdown (Fig. 4a). benzyloxycarbonyl-Val-Ala-Asp tripeptide with fluoromethyl group at the C-terminal) penetrates cells and reacts with yeast, mouse NGLY1 suppression upregulates transcription factors that activate and human NGLY115,42,43 by forming a covalent bond with the the expression of interferon λ genes in melanoma cells cysteine residues at the catalytic sites (e.g., Cys309 in human By mapping ~1400 manually curated, sequence-specific DNA- NGLY1), we docked Z-VAD-fmk to the human NGLY1 homology binding transcription factors38 to our global gene expression model using AutoDock. As expected, the top-scored binding profiling data, we discovered that the expression of 155 poses of Z-VAD-fmk in our homology model include the ones that transcriptional factor genes was altered (P < 0.05) due to NGLY1 are similar to the binding pose in the crystal structure of a mouse knockdown in melanoma cells. Among these differentially NGLY1 and Z-VAD-fmk complex (PDB code: 2F4O; Fig. 6b), expressed transcription factors, we found that the average RNA indicating that Z-VAD-fmk can bind to and inactivate the catalytic expression levels of IRF1, IRF3, IRF7, IRF9, REL (an NF-κB subunit) site of human NGLY1 in a similar fashion. This result also supports and RELB were elevated in the NGLY1-knockdown cells by 2.41, the accuracy and suitability of our NGLY1 homology model for 1.25, 2.23, 1.66, 1.12 and 1.65 folds, respectively. In addition, the inhibitor screening. expression and activation39,40 of IRF3 and IRF7 proteins was We designed four small molecules (NM-322, NM-348, NM-350 enhanced in melanoma cells with NGLY1 knockdown (Fig. 4d). and NM-354) that mimicked the N-acetylglucosamine (GlcNAc)- Since the expression of the interferon λ gene family members linked asparagine substrates of NGLY1 and contained strategically including the IL-29 (interferon λ1) gene is primarily regulated by positioned electrophilic groups (Fig. 6b). Their preferred binding the IRF proteins and NF-κB,41 it is likely that NGLY1 suppression poses in the human NGLY1 homology model included catalytic causes robust upregulation of IRF and NF-κB transcription factors poses where the electrophilic reactive moieties of compounds and subsequently upregulates IL-29 expression in melanoma cells. were oriented toward Cys309. The binding energy of these compounds in their most favourable catalytic poses ranged from NGLY1 suppression hinders melanoma tumour growth in vivo -6.3 to -6.9 kcal/mol. Since the binding energy of the most Using a xenograft tumour model (Fig. 5a), we test the anti- favourable binding pose of Z-VAD-fmk in the homology model melanoma response of NGLY1 suppression. The growth of was estimated as -6.3 kcal/mol, considering the ± 2.0 kcal/mol melanoma tumours established with SK-MEL-2 cells in mice was variation of binding energy calculation in AutoDock, it is impeded by the induced knockdown of NGLY1 (Fig. 5b, c). It was believable that our novel compounds could have comparable also noticed that three out of eight tumours with NGLY1-targeting inhibitory activity towards human NGLY1 as Z-VAD-fmk. Like Z- shRNA that increased their size during the initial 3-4 weeks of dox VAD-fmk and WRR139,15 our compounds caused NGLY1 inhibition treatment showed regression at the end of the study (Fig. 5b, and blocked the deglycosylation of denatured RNase B in vitro inset). The enhanced expression of GADD153, IRF3 and IL-29 was (Fig. 6c, upper panel). Since NGLY1 inactivation hinders the also detected in the NGLY1-knockdown tumours (Fig. 5c, d). These deglycosylation and proteolytic processing of NFE2L115 in findings attest to the in vivo antitumour efficacy of NGLY1 proteasome inhibitor-treated cells, we tested whether our novel inhibition that was expected from our in vitro studies. inhibitors may interfere with NFE2L1 deglycosylation. We used HEK293T cells (a sub-clone of HEK293 cells) in this test because Novel NGLY1 inhibitors show an anti-melanoma efficacy with a HEK293 cells have been used in a similar study and appear to limited impact on normal cells tolerate NGLY1 suppression well.15 HEK293T cells pretreated with Due to the lack of optimised small molecules that specifically NM-350 showed clear retention of N-glycans on NFE2L1, indicated inactivate human NGLY1, we aimed to develop novel NGLY1 by the electrophoretic mobility shift of full-length NFE2L1 (Fig. 6c, inhibitors and test their potential anti-melanoma use. Based on lower panel). The electrophoretic mobility shift of NFE2L1 due to the crystal structure of mouse NGLY1 (PDB code: 2F4M)42 as a glycan retention was also detected in NGLY1-deficient hPSCs model-building template, we built a homology model of the treated with bortezomib (Supplementary Figure S1D). human NGLY1 core domain (Fig. 6a) using a homology-modelling As shown in Fig. 6d, our novel compounds preferentially web server SWISS-MODEL. To test our human NGLY1 homology inhibited melanoma cell viability and had limited impact on model, we used it to perform in silico docking analysis of known normal cells. The treatment of our inhibitors also enhanced the Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1549 production and release of IFNβ and IL-29 in melanoma cells but not be directly caused by ER stress. Since ER stress caused by the not normal cells (Supplementary Figure S6A). Similar to the disruption of protein folding and ER function typically activates synergism between NGLY1 knockdown and dacarbazine treat- three arms of signalling pathways downstream of ER membrane- ment, NM-322 and dacarbazine caused a synergistic effect on the bound proteins that include PERK, ATF6, and IRE1α, we also suppression of SK-MEL-2 cells (Fig. 6e). Consistent with the examined the activation of these three signalling arms in NGLY1- potentiation of proteasome inhibitor cytotoxicity caused by knockdown melanoma cells. The enhanced phosphorylation of WRR139-mediated NGLY1 inhibition,15 NM-350 and bortezomib eIF2α was observed in melanoma cells due to the treatment of synergistically suppressed the viability of SK-MEL-2 cells (Fig. 6f). tunicamycin and NGLY1 knockdown (Supplementary Figure S2F), These results highlight that pharmacological inactivation of NGLY1 indicating the activation of the eIF2α signalling as the upstream of reduces melanoma cell viability and may be exploited for cancer ATF4. The activation of both ATF6 and XBP1, however, was not therapy purposes. seen in the NGLY1-knockdown cells (Supplementary Figure S2F). Interestingly, we detected the enhanced phosphorylation of PKR Melanoma cells present similar peptide signatures in response to and PERK in the NGLY1-knockdown cells (Supplementary Fig- NGLY1 knockdown and novel NGLY1 inhibitors ure S2G). These findings further support that the integrated stress Having elevated contents of GlcNAc-asparagine-containing pep- response signalling,44 rather than ER stress, could be a key tides and the differential abundance of specific proteins detected mediator for the NGLY1 suppression-induced activation of ATF4 in NGLY1-knockdown melanoma cells (Supplementary Figure S5; and GADD153 in melanoma cells. Supplementary Table S4), we examined proteomic changes in Through global gene expression profiling (Fig. 3), we melanoma cells treated with our NGLY1 inhibitors. Like the discovered that, in addition to the activation of stress- NGLY1-knockdown cells, cells treated with our inhibitors showed response signalling, NGLY1 suppression in melanoma cells relatively high contents of GlcNAc-asparagine-containing peptides causes the significant upregulation of interferon genes that (Supplementary Figure S6B), suggesting that our inhibitors are have well documented anticancer activity. Our data (Fig. 4) likely to suppress NGLY1 activity and allow ENGase to generate demonstrated that these cytokine surges play an important role more enzymatic products with the GlcNAc-asparagine signature. in melanoma cell death as the consequence of NGLY1 inhibition. Similar alteration patterns associated with the differentially NGLY1 suppression-stimulated cytokine responses also offer abundant proteins identified in the NGLY1-knockdown cells were direct evidence supporting the feasibility of immunomodula- observed in melanoma cells treated with NM-350 (Supplementary tion, particularly in malignant cells, by targeting NGLY1. Thus, Table S4), further supporting the NGLY1-inhibitory and anticancer NGLY1 suppression may potentiate the anticancer activity of activity of our novel compounds. immune-modulatory agents. A recent study has revealed that the chemical-induced activation of type I interferon (e.g., Ifnb1) signalling and immunogenic cell death potentiates the anti- DISCUSSION tumour efficacy of anti-PD-1 antibody in syngeneic mouse In this study, we revealed the significance and molecular tumour models of colon and bladder cancer.45 Since anti-PD-1 mechanisms of NGLY1 as a novel target in melanoma for the antibodies such as nivolumab and pembrolizumab are used for induction of cancer cell-specific apoptosis and the development of treating patients with melanoma and non–small cell lung new anticancer approaches. cancer, it is reasonable to consider the induction of cytokine From studying a genetic disorder known as NGLY1 deficiency, surges by targeting NGLY1 as a promising approach to enhance we realised that critical organs maintain their necessary functions the efficacy of immune checkpoint therapies in these patients. In for the vitality of NGLY1-deficient individuals.5,9 Although many the future, the immune responses triggered by NGLY1 inhibition types of cells present abnormal features in patients with NGLY1- may be exploited for antiviral purposes because the NGLY1 deficiency, these abnormalities may be largely attributed to suppression-activated cytokine signalling is also relevant to abnormal embryonic development rather than the direct effects of antiviral responses in cells.46 NGLY1 loss on terminally differentiated somatic cells. Thus, the We computationally modelled the conformations and interac- impact of NGLY1 suppression on differentiated somatic cells could tions of human NGLY1 with different potential inhibitors. Like our be quite mild. We indeed observed that NGLY1 suppression leads newly designed inhibitors, WRR139 showed a decent binding to limited perturbation in human normal cells while causing affinity with its electrophilic reactive moiety oriented towards detrimental outcomes in melanoma cells. Cys309 at the catalytic domain of human NGLY1 (Supplementary Through proteomic analysis, we not only observed a higher Figure S7A). Although WRR139 and our novel compounds may content of peptides containing GlcNAc-asparagine residues but both effectively bind to human NGLY1, their binding poses in also identified proteins showing altered abundance in melanoma NGLY1 were distinct (Supplementary Figure S7B). This difference cells with NGLY1 inhibition (Supplementary Table S4). However, in suggests that there could be sufficient space at the catalytic site of transcriptomic analysis, the genes that encode several of these NGLY1 to permit additional possibilities for compound derivation differentially abundant proteins (e.g., PPIA, VDAC1, PRKCSH, and optimisation to ultimately achieve a competitive inhibitor LASP1) did not appear differentially expressed at the RNA level with high specificity and other desired features for clinical in the NGLY1-inhibited cells, indicating that NGLY1 inhibition may applications. affect cell signalling networks by perturbing post-transcriptional Our results support a conclusion that human normal and or post-translational regulatory mechanisms. The links between melanoma cells present distinct features in response to NGLY1 and post-translational regulatory mechanisms were also NGLY1 suppression. In addition, NGLY1 suppression generates supported by the findings from several recent studies on multifaceted anticancer responses (Supplementary Figure S8) NGLY1.15,17–19 in vitro and in vivo. Our discoveries serve as a vivid proof for Although ATF4 and GADD153 were upregulated in NGLY1- the high value of understanding the biology hinted at by a rare knockdown melanoma cells, we did not observe a clear genetic disorder like NGLY1 deficiency. Reciprocally, our findings upregulation of ER chaperones GRP78/94 (Fig. 2c). In fact, the about NGLY1 suppression-caused cytokine responses in diseased abundance of GRP78/94 proteins in melanoma cells determined cells indicate that immunity alterations may occur in the NGLY1- by mass spectrometry appeared to drop in response to deficient patients whose development, health conditions, and NGLY1 suppression (Supplementary Table S4). The missing of susceptibility to disease or pathogens require close monitoring. GRP78/94 upregulation suggests that the NGLY1 suppression- We believe that continuous studies regarding NGLY1 function in induced activation of ATF4 and GADD153 in melanoma cells may normal development and pathological conditions are likely to Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1550 shed light on new strategies for managing multiple human 9. Lam, C. et al. Prospective phenotyping of NGLY1-CDDG, the first congenital disorders. disorder of deglycosylation. Genet. Med. 19, 160–168 (2016). 10. Uhlen, M. et al. A human protein atlas for normal and cancer tissues based on antibody proteomics. Mol. Cell. Proteom. 4, 1920–1932 (2005). ACKNOWLEDGEMENTS 11. Jones, J. C. et al. Melanocytes derived from transgene-free human induced pluripotent stem cells. J. Invest. Dermatol. 133, 2104–2108 (2013). We thank Dr. Rena Lapidus and her team members for their professional assistance in 12. Wang, Y. C. et al. Specific lectin biomarkers for isolation of human pluripotent the execution of the animal work in the translational core laboratory at the University stem cells identified through array-based glycomic analysis. Cell Res. 21, of Maryland, Baltimore. 1551–1563 (2011). 13. Wang, Y. C. et al. Targeting endoplasmic reticulum stress and Akt with OSU-03012 and gefitinib or erlotinib to overcome resistance to epidermal growth factor AUTHORS CONTRIBUTIONS receptor inhibitors. Cancer Res. 68, 2820–2830 (2008). A.Z. and V.J.T.L. contributed to study concept, data collection and analysis, and 14. Muller, F. J. et al. A bioinformatic assay for pluripotency in human cells. Nat. manuscript preparation. N.M.M. and K.A.E. contributed to the design, synthesis and Methods 8, 315–317 (2011). chemical characterisation of novel NGLY1 inhibitors and manuscript preparation. Y.Y. 15. Tomlin, F. M. et al. Inhibition of NGLY1 Inactivates the Transcription Factor Nrf1 and P.H. contributed to proteomic analysis and data interpretation. H.H., R.S. and J.L. and Potentiates Proteasome Inhibitor Cytotoxicity. ACS Cent. Sci. 3, 1143–1155 contributed to the computational modelling and analysis of human NGLY1 and its (2017). interactions with inhibitor candidates. A.P. contributed to data processing and 16. Wang, Y. C., Peterson, S. E. & Loring, J. F. Protein post-translational modifications statistical analysis. X.L. contributed to the histopathological analysis of NGLY1 and regulation of pluripotency in human stem cells. Cell Res. 24, 143–160 expression in patients’ tumour samples. Y.C.W. supervised the study, coordinated (2014). research efforts, and contributed to study concept and design, data analysis, and 17. Lehrbach N. J., Ruvkun G. Proteasome dysfunction triggers activation of SKN-1A/ manuscript writing. All authors reviewed and approved the manuscript. Nrf1 by the aspartic protease DDI-1. eLife 2016; 5: https://doi.org/10.7554/ eLife.17721. 18. Owings, K. G., Lowry, J. B., Bi, Y., Might, M. & Chow, C. Y. Transcriptome and ADDITIONAL INFORMATION functional analysis in a Drosophila model of NGLY1 deficiency provides insight Supplementary information is available for this paper at https://doi.org/10.1038/ into therapeutic approaches. Hum. Mol. Genet. 27, 1055–1066 (2018). s41416-018-0265-9. 19. Koizumi S., et al. The aspartyl protease DDI2 activates Nrf1 to compensate for proteasome dysfunction. eLife 2016; 5: https://doi.org/10.7554/eLife.18357. Competing interests: The authors declare no competing interests. 20. Luethy, J. D. & Holbrook, N. J. Activation of the gadd153 promoter by genotoxic agents: a rapid and specific response to DNA damage. Cancer Res. 52,5–10 (1992). Ethics approval and consent to participate: The use of hPSCs and vertebrate 21. Yang, H. et al. RG7204 (PLX4032), a selective BRAFV600E inhibitor, displays potent animals in this study was approved by the institutional committees. The detailed antitumor activity in preclinical melanoma models. Cancer Res. 70, 5518–5527 information was provided under Materials and Methods. (2010). 22. Anderson, D. J. et al. Targeting the AAA ATPase p97 as an approach to treat Consent for publication: All authors have read and approved the manuscript, its cancer through disruption of protein homeostasis. Cancer Cell. 28, 653–665 content, and its publication. (2015). 23. Kuo, T. F. et al. Protein disulfide isomerase a4 acts as a novel regulator of Availability of data and material: Upon appropriate request with the completion of cancer growth through the procaspase pathway. Oncogene 36, 5484–5496 interinstitutional material transferring agreements/licensing, the cell clones and novel (2017). compounds used in this study could be made available for other researchers’ use. 24. Wang, R. et al. Histone H4 expression is cooperatively maintained by IKKbeta and The data of global gene expression profiling have been deposited in an open access Akt1 which attenuates cisplatin-induced apoptosis through the DNA-PK/RIP1/ database. The detailed information was provided under Materials and Methods. IAPs signaling cascade. Sci. Rep. 7, 41715 (2017). 25. Cerezo, M. et al. Compounds triggering ER stress exert anti-melanoma effects and overcome BRAF inhibitor resistance. Cancer Cell. 29, 805–819 (2016). Funding: This work is supported by the UNTHSC Start-up Fund for Stem Cell 26. Bidwell, B. N. et al. Silencing of Irf7 pathways in breast cancer cells promotes Laboratory and Faculty Pilot Grant FY15 (RI6182) to Y.C.W, Start-up Fund for Medicinal bone metastasis through immune escape. Nat. Med. 18, 1224–1231 (2012). Chemistry Laboratory to K.A.E., and Start-up Fund for Computational Chemistry 27. Reu, F. J. et al. Overcoming resistance to interferon-induced apoptosis of renal Laboratory to J.L. V.J.T.L. is supported by the NIH NRSA Institutional Predoctoral carcinoma and melanoma cells by DNA demethylation. J. Clin. Oncol. 24, Training Grant (T32 AG020494), and predoctoral fellowships from the American 3771–3779 (2006). Medical Association (AMA) Foundation and the Cancer Prevention Research Institute 28. Vert, A., Castro, J., Ribo, M., Benito, A. & Vilanova, M. Activating transcription factor of Texas (CPRIT; RP170301). Y.Y.H. and P.H. are supported by the Ovarian Cancer 3 is crucial for antitumor activity and to strengthen the antiviral properties of Research Foundation. Onconase. Oncotarget 8, 11692–11707 (2017). 29. Hassan, M. et al. The BH3-only member Noxa causes apoptosis in melanoma cells – REFERENCES by multiple pathways. Oncogene 27, 4557 4568 (2008). 30. Ishiguro, H. et al. Identification of AXUD1, a novel human gene induced by AXIN1 1. Huang, C. et al. Endo-beta-N-acetylglucosaminidase forms N-GlcNAc protein and its reduced expression in human carcinomas of the lung, liver, colon and aggregates during ER-associated degradation in Ngly1-defective cells. Proc. Natl. kidney. Oncogene 20, 5062–5066 (2001). Acad. Sci. USA 112, 1398–1403 (2015). 31. Jalili, A. et al. Dual suppression of the cyclin-dependent kinase inhibitors 2. Caglayan, A. O. et al. NGLY1 mutation causes neuromotor impairment, intellec- CDKN2C and CDKN1A in human melanoma. J. Natl. Cancer Inst. 104, 1673–1679 tual disability, and neuropathy. Eur. J. Med. Genet. 58,39–43 (2015). (2012). 3. Need, A. C. et al. Clinical application of exome sequencing in undiagnosed 32. Falkenius, J. et al. High expression of glycolytic and pigment proteins is asso- genetic conditions. J. Med. Genet. 49, 353–361 (2012). ciated with worse clinical outcome in stage III melanoma. Melanoma Res. 23, 4. Suzuki, T. The cytoplasmic peptide:N-glycanase (Ngly1)-basic science encounters 452–460 (2013). a human genetic disorder. J. Biochem. 157,23–34 (2015). 33. Goplen, D. et al. alphaB-crystallin is elevated in highly infiltrative apoptosis- 5. Enns, G. M. et al. Mutations in NGLY1 cause an inherited disorder of the endo- resistant glioblastoma cells. Am. J. Pathol. 177, 1618–1628 (2010). plasmic reticulum-associated degradation pathway. Genet. Med. 16, 751–758 34. Slipicevic, A. et al. The fatty acid binding protein 7 (FABP7) is involved in pro- (2014). liferation and invasion of melanoma cells. Bmc. Cancer 8, 276 (2008). 6. Funakoshi, Y. et al. Evidence for an essential deglycosylation-independent activity 35. Guenterberg, K. D. et al. Interleukin-29 binds to melanoma cells inducing Jak- of PNGase in Drosophila melanogaster. PLoS ONE 5, e10545 (2010). STAT signal transduction and apoptosis. Mol. Cancer Ther. 9, 510–520 (2010). 7. Maerz, S., Funakoshi, Y., Negishi, Y., Suzuki, T. & Seiler, S. The Neurospora peptide: 36. Jablonska, J., Leschner, S., Westphal, K., Lienenklaus, S. & Weiss, S. Neutrophils N-glycanase ortholog PNG1 is essential for cell polarity despite its lack of enzy- responsive to endogenous IFN-beta regulate tumor angiogenesis and growth in matic activity. J. Biol. Chem. 285, 2326–2332 (2010). a mouse tumor model. J. Clin. Invest. 120, 1151–1164 (2010). 8. Heeley, J. & Shinawi, M. Multi-systemic involvement in NGLY1-related disorder 37. Lau, E. et al. Transcriptional repression of IFNbeta1 by ATF2 confers melanoma – caused by two novel mutations. Am. J. Med. Genet. A. 167, 816 820 (2015). resistance to therapy. Oncogene 34, 5739–5748 (2015). Stress and interferon signalling-mediated apoptosis contributes to. . . A Zolekar et al. 1551 38. Vaquerizas, J. M., Kummerfeld, S. K., Teichmann, S. A. & Luscombe, N. M. A census 45. Hossain, D. M. S. et al. Dinaciclib induces immunogenic cell death and enhances of human transcription factors: function, expression and evolution. Nat. Rev. anti-PD1-mediated tumor suppression. J. Clin. Invest. 128, 644–654 (2018). Genet. 10, 252–263 (2009). 46. Ivashkiv, L. B. & Donlin, L. T. Regulation of type I interferon responses. Nat. Rev. 39. Lin, R., Heylbroeck, C., Pitha, P. M. & Hiscott, J. Virus-dependent phosphorylation Immunol. 14,36–49 (2014). of the IRF-3 transcription factor regulates nuclear translocation, transactivation potential, and proteasome-mediated degradation. Mol. Cell. Biol. 18, 2986–2996 (1998). Open Access This article is licensed under a Creative Commons 40. Sharma, S. tenOever BR, Grandvaux N, Zhou GP, Lin R, Hiscott J. Triggering the Attribution 4.0 International License, which permits use, sharing, interferon antiviral response through an IKK-related pathway. Science 300, adaptation, distribution and reproduction in any medium or format, as long as you give 1148–1151 (2003). appropriate credit to the original author(s) and the source, provide a link to the Creative 41. Osterlund, P. I., Pietila, T. E., Veckman, V., Kotenko, S. V. & Julkunen, I. IFN reg- Commons license, and indicate if changes were made. The images or other third party ulatory factor family members differentially regulate the expression of type III IFN material in this article are included in the article’s Creative Commons license, unless (IFN-lambda) genes. J. Immunol. 179, 3434–3442 (2007). indicated otherwise in a credit line to the material. If material is not included in the 42. Zhao, G. et al. Structure of the mouse peptide N-glycanase-HR23 complex sug- article’s Creative Commons license and your intended use is not permitted by statutory gests co-evolution of the endoplasmic reticulum-associated degradation and regulation or exceeds the permitted use, you will need to obtain permission directly DNA repair pathways. J. Biol. Chem. 281, 13751–13761 (2006). from the copyright holder. To view a copy of this license, visit http://creativecommons. 43. Misaghi, S., Pacold, M. E., Blom, D., Ploegh, H. L. & Korbel, G. A. Using a small org/licenses/by/4.0/. molecule inhibitor of peptide: N-glycanase to probe its role in glycoprotein turnover. Chem. Biol. 11, 1677–1687 (2004). 44. Pakos-Zebrucka, K. et al. The integrated stress response. EMBO Rep. 17, © The Author(s) 2018 1374–1395 (2016).